Will AI leave us in charge of our own work? - Economics Observatory

TEXT START: Although artificial intelligence can raise workers’ productivity and leave them free from interference, it can still erode their autonomy.

The Dissection

The text identifies a real secondary casualty of automation: workers can retain formal authority while losing substantive authorship. “Human oversight” can become liability assignment—humans remain answerable for machine-produced outputs while contributing little to them.

But the article confines the autopsy to job quality. It treats the workplace as a design problem: preserve judgment-rich tasks, involve workers, and align accountability with authorship. That is a local repair applied to a systemic rupture. It asks how workers can remain meaningful participants after AI absorbs the work, rather than whether they will retain access to economically necessary work at all.

The Core Fallacy

The central error is confusing residual participation with durable economic power.

Under the Discontinuity Thesis, AI does not merely deskill or hollow out jobs. It attacks the mass employment → wage → consumption circuit. Once AI can perform cognitive work more cheaply and effectively, “authorship” becomes a privilege allocated by owners of AI capital, not a stable right secured by policy.

The article’s preferred solution—designing jobs so humans retain judgment-rich tasks—depends on firms voluntarily preserving costly human contribution when automated systems can produce the result without it. Competitive pressure eventually strips away that subsidy. The worker may remain the nominal author, but only as a ceremonial signatory attached to a machine’s output.

The article correctly sees that human-in-the-loop rules can create responsibility without control. It fails to follow that logic to its conclusion: the worker’s deeper problem is not merely loss of standing inside a job. It is loss of bargaining power, income, and eventually the job’s economic necessity.

Hidden Assumptions

• AI will remain complementary enough that firms continue needing large numbers of human judgment-makers.
• Organisations can preserve human-authored work without losing to competitors that automate it fully.
• Workers and representatives possess enough power to shape deployment decisions.
• Regulation can stabilize human-only or human-dominant economic domains at scale.
• Accountability can be fairly matched to authorship without changing ownership of the systems producing the work.
• “Good jobs” remain the relevant social objective after productive participation collapses.
• Formal employment will continue to distribute status, income, and social membership.
• Human judgment has durable market value independent of whether AI can replicate its outputs.

These are not minor omissions. They are the load-bearing beams of the article’s policy case.

Social Function

Classification: partial truth, transition management, and ideological anesthetic.

The partial truth is valuable: hollow oversight is not autonomy, and accountability without authorship is an institutional trap.

The transition-management function is to make automation politically acceptable by promising redesigned, dignified human roles around increasingly autonomous systems. The ideological anesthetic is subtler: it relocates the conflict from ownership and distribution to workplace architecture. It invites policymakers to preserve the appearance of meaningful participation while leaving control of AI capital untouched.

The likely endpoint is not the autonomous worker but the accountable chaperone: a human retained to validate, explain, absorb blame, and reassure customers while the machine performs the substance of the work.

The Verdict

The article diagnoses the corpse’s loss of dignity while avoiding the cause of death.

Its concept of authorship is analytically sharp at the task level, but economically fragile at the system level. If AI reaches durable superiority across cognitive work and institutions cannot preserve human-only domains, authorship will not remain a general feature of employment. It will become a scarce privilege for Sovereigns and indispensable Servitors, while the majority are pushed into verification, liability management, maintenance, logistics, care, or nonproductive transfer dependence.

“Keep a human in charge” is indeed the wrong test. The harsher test is: who owns the system, who receives the surplus, and whether the human is still economically necessary. The article answers none of those questions. That omission is not peripheral. It is the mechanism by which a terminal labor system is cosmetically refurbished before abandonment.